EU AI Act Transparency: Which Customer-Facing Outputs Need Review Before December?

By DataTip · Published November 7, 2026

For customer-facing AI, the first decision is what kind of output or interaction you need to assess. The EU AI Act transparency date to check for watermarking AI-generated content is 2 December 2026. That is not the only relevant date: separate Article 50 disclosure requirements have a stated application date of 2 August 2026 for AI systems placed on the market after that date.

The Digital Omnibus sets 2 December 2026 as the date watermarking obligations for AI-generated content apply. Separate Article 50(1) AI-interaction and Article 50(4) deepfake disclosure requirements have a stated application date of 2 August 2026 for AI systems placed on the market after that date. Confirm output type, system scope and organisational role against authoritative guidance before selecting controls.

The European Commission adopted its final guidelines on Article 50 transparency obligations on 20 July 2026. The Digital Omnibus, EU Regulation 1744/2026, became applicable on 27 July 2026. Mayer Brown’s legal update identifies the Digital Omnibus as EU Regulation 1744/2026 and reports its publication in the Official Journal. The update also covers the Commission’s guidance on Article 50 transparency obligations.

What does EU AI Act transparency require before December 2026?

EU AI Act transparency does not have one start date for every requirement. The Digital Omnibus sets 2 December 2026 for watermarking obligations on AI-generated content. Separate Article 50(1) and Article 50(4) disclosure requirements have a stated 2 August 2026 application date for AI systems placed on the market after that date.

The Omnibus changes some deadlines, but it does not delay all transparency requirements. Article 50(1) concerns disclosure that a person is interacting with AI; Article 50(4) concerns disclosure of deepfakes. Neither should be collapsed into the separate watermarking date.

The dates do not, on their own, establish whether a particular system or output is covered. The customer-facing review should start with the output or interaction, then establish which system is involved and the organisation’s role. This avoids treating a general summary of dates as a case-specific applicability decision.

Which customer-facing outputs should you review?

Review customer-facing situations by type rather than treating all AI use as one category. AI-generated content raises the watermarking question; an interaction with AI raises the Article 50(1) disclosure question; and deepfakes raise the separate Article 50(4) disclosure question. The relevant requirement and scope still need confirmation.
A practical review can identify:

• AI-generated content that may fall within the watermarking obligation;
• interactions where a person may need to be told they are interacting with AI; and
• outputs that may raise the deepfake disclosure requirement.
  These categories help structure the inquiry; they are not findings that every output in a category is covered. In particular, the stated 2 August 2026 date for Article 50(1) and Article 50(4) concerns AI systems placed on the market after that date. Keep that qualification attached to the date when recording the review’s scope.

How should you assess system scope and organisational role?

Before choosing controls, identify the customer-facing output or interaction, the AI system involved, and your organisation’s relevant role. Then confirm whether the applicable transparency requirement covers that combination. The source does not establish identical obligations for every organisation, system, or use.

Providers and deployers are among the organisations that may need to assess their roles, but the source does not determine the position for a particular organisation. Do not infer applicability from the fact that content reaches a customer alone. The output type, system scope, market-placement condition where relevant, and organisational role all matter to the assessment.

Keep the review focused on those questions. Record what the customer encounters, whether the issue concerns generated content, an AI interaction, or a deepfake, which system is involved, and the role your organisation has in relation to it. Confirm the answer against authoritative guidance before deciding which controls are appropriate.

Are the later high-risk AI deadlines the same as transparency dates?

No. The later high-risk dates are separate from the transparency dates. The source gives 2 December 2027 for high-risk AI systems used in high-risk use cases, and 2 August 2028 for high-risk AI systems used as safety components in products covered by EU sectoral legislation.

For high-risk systems already in service, the revised dates apply only if those systems undergo significant design changes from the relevant date. The source also says the position for public-authority operators is unclear. These qualifications concern the high-risk timetable; they do not postpone the separate transparency requirements.

What should you confirm before choosing controls?

Confirm the output type, the relevant system and your organisation’s role before selecting controls. Check whether the stated requirement applies to that situation against authoritative guidance, including the market-placement qualification attached to the 2 August 2026 date where relevant. The dates alone are not an applicability assessment.

The next decision is therefore scoped: determine which transparency question is raised by the customer-facing output, and confirm whether it applies to your system and role. Do not assume that one deadline or one obligation covers every use.

Key takeaways

• The 2 December 2026 date concerns watermarking obligations for AI-generated content, not every transparency requirement.
• The stated 2 August 2026 date for Article 50(1) and Article 50(4) concerns AI systems placed on the market after that date.
• AI-generated content, AI interactions and deepfakes raise distinct transparency questions.
• Confirm system scope and organisational role against authoritative guidance before choosing controls.
• The later high-risk AI dates are separate from the transparency dates.

Practical tips

• Classify the customer-facing situation as generated content, an AI interaction, or a potential deepfake.
• Identify the AI system and the organisation’s role in relation to it.
• Check the market-placement condition and confirm applicability against authoritative guidance before selecting controls.